One of the law’s key aspects is the funding it provides to the National Institutes of Health, which would amount to $4.8 billion over 10 years. “There are a lot of things that sound good about the law,” says Michael Repka, MD, the American Academy of Ophthalmology’s medical director for government affairs, and a professor of ophthalmology and pediatrics at Johns Hopkins in Baltimore. “The $4.8 billion or so would be a terrific improvement in a space that’s been somewhat stifled in recent years [out of] particular concern about too much spending among members of Congress and even on the executive side.
“[Another] thing that’s probably good for us involves lessening EHR-blocking regulations,” Dr. Repka continues. “We have had issues with IRIS registry and data blocking by certain large EMR vendors. In order for [AAO] members to be successful in the new quality payment program, particularly the [Merit-based Incentive Payment System], it’s really important for us to have access there. So, we hope that will get better with this law. Specifically, many of the AAO members who aren’t in the IRIS registry happen to have one of the large, institutional-based EMR’s such as EPIC, which, for the most part, can’t participate in the IRIS registry. Of course, participating in the IRIS registry is important because it’s a relatively painless way to succeed in MIPS.”
There’s a chance, however, that the billions in funds might not materialize. “The biggest promise of the law—greater funding for the NIH—is a bit illusory, in that it’s subject to the appropriations process each year,” cautions Ameet Sarpatwari, PhD, JD, Instructor in Medicine at Harvard Medical School and assistant director of the Program on Regulation, Therapeutics, and Law at Brigham and Women’s Hospital in Boston. “And, as we’ve seen in years past, that sort of set-aside can be raided during that process, so it’s not necessarily a sure thing.”
Dr. Sarpatwari goes on to outline other aspects of the law that could have some negative effects down the road depending on how they’re implemented. “One concern regards supplemental approval of drugs, meaning drugs that are already on the market but are looking to gain another foothold in terms of another indication,” he says. “The bill potentially lowers the bar for their approval while, at the same time, purporting not to. It essentially commands the FDA commisioner to promulgate new guidance on the use of the various ways to get a drug approved, which, in this case, is the use of real-world evidence. This provides a bit more cover to allow drugs with less-clear evidence of efficacy to go through. Essentially, it’s saying, ‘How can we use just regular, observational data to support a supplemental drug approval?’ That’s problematic because that type of evidence can be very misleading, and isn’t subject to the same rigor that a randomized clinical trial is.
“The second area involves the issue of medical devices,” Dr. Sapartwari continues. “There’s an existing pilot program that would allow so-called breakthrough devices to be approved under a sort of accelerated time frame using, basically, more questionable evidence. What the Cures Act would do is formalize this program and expand it to potentially allow more products under it, because the products that can be classified as ‘breakthrough’ under this pathway don’t necessarily have to be clinically meaningful. So, there’s an inherent tension if you speed up the review of products and base it on less-rigorous evidence. You’re going to allow more devices through that aren’t, again, necessarily that effective. And, if they aren’t that effective, this changes the risk/benefit profile of these products, and results in significant waste for patients and taxpayers who have to pay for them. This is important in the context that, when you look at the evidence base that supports device approvals, it is already far less than that which supports new drugs.”
The third aspect that concerns some is the law’s use of the limited population pathway for antibiotic approval. “This would allow antibiotics to be approved on a sliding benefit/risk scale,” Dr. Sarpatwari explains. “And so, you can imagine again that it allows the FDA commissioner to use his or her discretion to steer through more products that previously wouldn’t have gotten through. Then, the law would require the product to indicate that it had been approved under this limited pathway. However, we know that, based on evidence, particularly from the nutritional supplement industry, consumers don’t really heed these disclaimers very well.”
Dr. Repka says worries over lax approval standards have always accompanied FDA revamps. “That’s always been the dilemma with the FDA,” he says. “Where do you put the cut point between safety, aka patient protections, and technology advancement? That’s tough to do. This seems to lean both ways: improving the support of the safety side but also helping to promote some innovation. I think that anytime the FDA gets pushed to move things along, this is going to be a very appropriate concern. In no way should we downplay that, and we should continue to say that the agency, as it implements the programs as designed, shouldn’t pay short shrift to the approval process.”
Google Algorithm Detects Diabetic Retinopathy
In a study sponsored by Google, researchers tested a newly developed deep learning computer algorithm designed to detect diabetic retinopathy and diabetic macular edema from retinal fundus photographs. Deep learning is a computational method which allows an algorithm to program itself through “learning.” The project’s system, called DeepMind, “learns” by studying a large set of examples that demonstrate the desired behavior and then adapting itself in response.
The algorithm graded 128,175 retinal images three to seven times each for diabetic retinopathy and DME. The images were also examined by a panel of 54 U.S.-licensed ophthalmologists. When set for high specificity, in two validation sets composed of 9,963 images and 1,748 images, the algorithm had a 90.3-percent and 87-percent sensitivity and 98.1-percent and 98.5-percent specificity, respectively, for detecting referable diabetic retinopathy—which was defined as moderate or worse diabetic retinopathy—or referable macular edema by the majority decision of a panel of at least seven of the ophthalmologists. When set for high sensitivity, the algorithm had 97.5-percent and 96.1-percent sensitivity and 93.4-percent and 93.9-percent specificity, respectively, in the two validation sets.
In reviewing the data, one of the DeepMind researchers, Google’s Lily Peng, MD, PhD, observes, “The results show that our algorithm’s performance is on-par with that of ophthalmologists. For example, on the first validation set, the algorithm has an F-score [combined sensitivity and specificity metric] of 0.95, which is slightly better than the median F-score of 0.91 achieved by the eight ophthalmologists we consulted. The significance here is that deep-learning algorithms had high sensitivity and specificity for detecting diabetic retinopathy and diabetic macular edema in retinal fundus photographs.”
With regard to the algorithm’s place in a clinical setting, its practical application requires further research. “Another open question is whether the design of the user interface and the online setting for grading used by ophthalmologists has any impact on their performance relative to a clinical setting,” says Dr. Peng. “Addressing this will require further experiments. The algorithm has only been trained to identify diabetic retinopathy and diabetic macular edema. It may miss non-diabetic retinopathy lesions that it was not trained to identify. Hence, this algorithm is not a replacement for a comprehensive eye exam, as it will ignore components such as visual acuity, refraction, eye pressure measurements, etc. However, with further research, the results suggest that the algorithm could lead to improved care and outcomes compared with the current ophthalmologic assessment.”
Although further research is necessary, this new deep-learning algorithm has potential use in telemedicine, as it will allow patients to “self-diagnose” in the comfort of their own homes, even if it is only for diabetic retinopathy and diabetic macular edema. “Along with telemedicine, technologies such as these could increase access to care and assist in screening for diabetic retinopathy in areas with limited resources,” Dr. Peng claims.
“Despite these exciting findings, there are a few limitations to this system,” Dr. Peng notes. “The algorithm may not perform as well for images with subtle findings that a majority of ophthalmologists would miss. Another limitation arises from the nature of deep networks, in which the neural network was provided with only the image and associated grade, without explicit definitions of features (microaneurysms, exudates). Because the network ‘learned’ the features that were most predictive, it is indeed possible that the algorithm is using features previously unknown to or ignored by humans. Understanding which features are used to make predictions will be an important area of investigation for further studies, and is indeed an active area of research within the larger machine-learning community.”
Fovista Fails to Meet Phase III Endpoint
Hot on the heels of apparently positive data from its Phase IIb trial, the platelet-derived growth factor drug Fovista (Ophthotech) didn’t provide the visual acuity benefit that was aimed for in its pivotal Phase III study.
The Phase III Fovista clinical trials, OPH1002 and OPH1003, were international, multicenter, randomized, double-masked, controlled studies evaluating 1.5 mg of Fovista administered in combination with Lucentis, an approach dubbed “Fovista combination therapy,” compared to Lucentis injection alone, for the treatment of wet age-related macular degeneration. In each trial, patients were randomized to one of two approximately equal sized treatment groups. The two Phase III trials enrolled 1,248 patients with wet AMD.
In data released by Ophthotech, patients receiving Fovista combo therapy gained a mean of 10.24 letters at 12 months compared to a mean gain of 10.01 letters for patients receiving Lucentis monotherapy. In OPH1002, consisting of 619 patients, combo-therapy patients gained a mean of 10.74 letters at 12 months, compared to a mean gain of 9.82 ETDRS letters in Lucentis-only patients. In OPH1003, consisting of 626 treated patients, subjects receiving combo therapy gained a mean of 9.91 letters at 12 months vs. a mean gain of 10.36 ETDRS letters in patients receiving Lucentis monotherapy. None of these results of the pre-specified primary efficacy analysis were statistically significant, the researchers say.
In a pooled data analysis, 12.1 percent of patients receiving combo therapy lost five or more letters from baseline at one year, compared to 11.2 percent of patients receiving Lucentis alone. In OPH1002, 12 percent of patients receiving the combination lost five or more letters at month 12, compared to 12.3 percent of patients receiving Lucentis. And in OPH1003, 12.2 percent of patients receiving the Fovista combination lost five or more letters at a year, compared to 10.2 percent of the Lucentis patients.
Samir Patel, president of Ophthotech, expressed his disappointment in the results in a conference call following the release of the data. “We are in the process of analyzing the data in order to better understand this outcome,” he said. “We are most disappointed by these results—especially for patients.” REVIEW